Audience: AI coding agents first.

Perform repository launch-readiness JUDGEMENT sweep. READ-ONLY: NEVER modify, create, or commit files. Print stdout only.

OWNED IDS: AC-12, AC-13, AC-14, AC-15, AC-16, AC-20, AC-21, AC-22, AC-23, AC-24, AC-26, AC-28, AC-29, AC-30, AC-31

Checklist:
1. Return each owned ID exactly once. NEVER evaluate any other criterion.
2. Re-derive current-tree verdicts. NEVER inherit `GOLIVE.md` verdicts.
3. For AC-12, AC-13, AC-16, AC-24: open covering tests. WORKS requires durable outcome assertion: persisted row, API truth, or confirmation surviving reload.
4. For AC-14, AC-15, AC-20, AC-21, AC-22, AC-23, AC-26, AC-28, AC-29, AC-30: code/config existence is insufficient. Keep UNVERIFIED without dated recorded evidence of real-world behavior.
5. For AC-31: inspect current implementation and evidence. WORKS requires observed behavior or durable-outcome test.
6. Cite every evidence claim as `path/to/file.ext:LINE` using opened, resolvable lines.
7. For non-WORKS, name exact evidence or change required in `whatWouldSettleIt`.

Output one JSON array; print nothing after it:
[{ "id": "AC-12", "verdict": "works|broken|missing|unverified", "evidence": "<one to three cited sentences>", "whatWouldSettleIt": "<required when not works>" }]